Pauli equation with non-Hermitian PT -symmetric boundary...
Transcript of Pauli equation with non-Hermitian PT -symmetric boundary...
Czech Technical University in Prague
Faculty of Nuclear Sciences and Physical Engineering
Department of Physics
Pauli equation with non-HermitianPT -symmetric boundary condition
BACHELOR’S THESIS
Author: Radek Novak
Supervisor: Mgr. David Krejcirık, Ph.D.
Academic year: 2010/2011
Prohlasuji, ze jsem tuto bakalarskou praci vypracoval samostatne a vyhradne s pouzitım
uvedene literatury.
Nemam zavazny duvod proti pouzitı tohoto skolnıho dıla ve smyslu §60 Zakona
c.121/2000 Sb., o pravu autorskem, o pravech souvisejıcıch s pravem autorskym a o
zmene nekterych zakonu (autorsky zakon).
V Praze dne Radek Novak
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowlegments
I thank my advisor David Krejcirık for bringing me to the fascinating world of func-
tional analysis. I also thank my consultant Petr Siegl for his useful hints which helped
me get more insight into the topic.
Title: Pauli equation with non-Hermitian PT -symmetric boundary
condition
Author: Radek Novak
Advisor: Mgr. David Krejcirık, Ph.D.
Consultant: Ing. Petr Siegl
Abstract:
The Pauli Hamiltonian in a bounded interval subjected to arbitrary PT -symmetric
Robin-type boundary conditions is investigated. It is introduced as an m-sectorial
operator in a Hilbert space via quadratic form. We find the adjoint operator and con-
ditions on various symmetries of the Hamiltonian using the quadratic form approach.
We perform spectral analysis of the Hamiltonian and derive an implicit equation for
the eigenvalues. These results are applied together with numerical analysis on sev-
eral examples of boundary conditions and the metric operator is found for one of
these examples.
Keywords: Pauli equation, Robin boundary conditions, scattering, PT-symmetry,
spectral analysis, metric operator, quadratic form
Nazev: Pauliho rovnice s nehermitovskymi PT -symetrickymi
hranicnımi podmınkami
Autor: Radek Novak
Abstrakt:
V teto praci je zkouman Pauliho Hamiltonian na omezenem intervalu s obecnymi
PT -symetrickymi Robinovymi hranicnımi podmınkami. Je zaveden jako m-
sektorialnı operator na Hilbertove prostoru pomocı kvadraticke formy. Prıstupem
pres kvadratickou formu nalezame sdruzeny operator a podmınky ruznych symetriı
Hamiltonianu. Provadıme spektralnı analyzu Hamiltonianu a odvozujeme implicitnı
rovnici pro vlastnı cısla. Tyto vysledky pak aplikujeme spolu s numerickou analyzou
na nekolik prıkladu hranicnıch podmınek a nalezame metricky operator pro jeden z
techto prıkladu.
Klıcova slova: Pauliho rovnice, Robinovy hranicnı podmınky, PT-symetrie,
spektralnı analyza, metrika, kvadraticka forma
4
Contents
1 Introduction 6
2 PT -symmetry 8
2.1 Antilinear symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Pseudo-Hermiticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 C-self-adjointness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Physical realisation 14
4 Elements of the theory of sectorial forms 18
4.1 Sectorial forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Sobolev spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5 Definition of the Hamiltonian 24
5.1 Definition via quadratic form . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.2 Adjoint operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6 Spectral analysis 33
6.1 Examples of boundary conditions . . . . . . . . . . . . . . . . . . . . . . . 35
6.2 Metric operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7 Conclusions 42
A Wolfram Mathematica source code 43
Bibliography 45
5
Chapter 1
Introduction
Quantum mechanics is based, like other physical theories, on several axioms. One of these
axioms states that the physical observables correspond to linear operators acting in a
Hilbert space. It is also natural to require from these operators the reality of the points
of the spectrum since these are the possible outcomes of a measurement. Traditionally,
this is summarized in the condition that these operators have to be self-adjoint1. In the
case of the Hamiltonian, this condition also ensures the unitary (probability preserving)
time evolution of the system but it is not necessary for the reality of the spectrum.
The non-self-adjoint operators, which are self-adjoint in a Hilbert space with a different
scalar product, exist and this enables the creation of new alternative representations of
quantum mechanics.
One of these attempts to use non-self-adjoint operators in quantum mechanics is the
so-called PT -symmetric quantum mechanics. Its main concern are the Hamiltonians
exhibiting the simultaneous parity and time reversal symmetry. The interest in these
Hamiltonians arose from the observation that many of them possess a real spectrum.
Many such models were studied and it turned out that the key to the correct interpreta-
tion of these quantum systems is the finding of a bounded positive self-adjoint operator
Θ, which can be used to define the new scalar product. However, this is in general a dif-
ficult task. In many models the metric operator cannot be found explicitly, the majority
of known results provide just formal perturbative expansions. Therefore Krejcirık, Bıla
and Znojil developed in [21] a simple PT -symmetric model where they found an explicit
formula for the metric.
1We shall use in this text term self-adjoint which used usually in mathematical literature instead of
its counterpart hermitian used in physical papers.
6
In this thesis we are interested in the role of spin in the model. It extends the model to a
two dimensional matrix Hamiltonian and brings many classes of possible PT -symmetric
boundary conditions. An interesting feature of this extended model is its physical ap-
plication - the scattering of a charged particle in a magnetic field can be described via
this spectral problem [17].
This thesis is organized as follows. In the following Chapter 2 we bring a short overview
of possible approaches to the PT -symmetric quantum mechanics. In Chapter 3 we step
by step approach the Hamiltonian, which is the main concern of this thesis. A special
attention is paid to its scattering interpretation. Chapter 4 summarizes notions and
theorems of spectral theory used in this thesis. It also brings a brief introduction to the
theory of Sobolev spaces. Chapter 5 is dedicated to a rigorous mathematical definition
of the Hamiltonian via quadratic form and finding the adjoint operator. The spectral
analysis of the Hamiltonian is carried out in Chapter 6 and few examples of boundary
conditions are studied. This section includes the numerical analysis of these examples.
We conclude this thesis in Chapter 7, where we sum up our results and discuss their
possible future extensions.
7
Chapter 2
PT -symmetry
The idea of PT -symmetry first emerged with the recognition that many non-self-
adjoint Hamiltonians possess real spectrum. The family of Hamiltonians H = −∆ +
x2(ix)ε, ε ∈ R+ acting in L2(R) was investigated firstly and it was discovered numeri-
cally that the eigenvalues of these Hamiltonians are entirely real, positive, and discrete
[5]. This strange behaviour was attributed to the so-called PT -symmetry of the Hamilto-
nian. This symmetry is often physically interpreted as parity-time reflection symmetry,
however, there is deeper and more general mathematical background in this theory,
which was developed during recent years. In this chapter we will take a closer look on
three main approaches to the PT -symmetry.
To avoid confusion, which could arise, we introduce a notion of D-self-adjointness. We
say that an operator T is D-self-adjoint if for its adjoint operator holds
T ∗ = DTD (2.1)
for some linear or antilinear operator D. This term is in literature used for a variety of
operators D with different properties. It reduces in this text on the fulfillment of (2.1)
and it will be always stressed out which properties of the operator D are required.
2.1 Antilinear symmetry
First way how to deal with PT -symmetry is to introduce a general concept of the
antilinear symmetry. This concept allows among others to work with the PT -symmetric
models in a general Hilbert space H .
8
Definition 2.1.1. Let T be a densely defined closed operator in H . We say that T has
an antilinear symmetry if there exists an antilinear bijective operator S and the relation
TSψ = STψ (2.2)
holds for all ψ ∈ D(T ).
In the framework of PT -symmetric quantum mechanics the operator PT is chosen as
the antilinear symmetry S. The operator PT is composed of two parts - the parity P,
representing spacial symmetry and acting in L2(R) space
(Pψ)(x) := ψ(−x),
and the time-reversal T , standing for a complex conjugation
(T ψ)(x) := ψ(x).
Link between the time-reversal and the complex conjugation can be seen from the time-
dependent Schrodinger equation
Hψ = i~∂ψ
∂t. (2.3)
If we demand the system to be symmetric with respect to the time-reversal, the equation
(2.3) should not change after the swap of t for −t. But the right-hand side changes the
sign so we have to consider the effect of the complex conjugation operator to reproduce
the form of the Schrodinger equation.
Both P and T satisfy relation P2 = I, T 2 = I and clearly commute with each other.
Consequently, the operator PT is antilinear (since P is linear and T antilinear) and sat-
isfies (PT )2 = I. The property of PT -symmetry of the Hamiltonian is then understood
as fulfillment of the equation (2.2) for S = PT .
Let us investigate further properties of the PT operator, in particular its spectrum
and the influence it has on the spectrum of the Hamiltonian. We solve the equation
PT ψ = λψ. (2.4)
The equation (2.4) is transferred by application of operator PT and using the antilin-
earity of PT into the form ψ = |λ|ψ. Therefore the eigenvalues of PT , λ = eiφ, are pure
phases.
9
Assuming that E is an eigenvalue of the PT -symmetric Hamiltonian H associated
with an eigenvector ψE , we get, after simple algebraic manipulation,
HPT ψE = EPT ψE ,
so E is also an eigenvalue of H, associated with an eigenvector PT ψE . The eigenvalue
E is real if and only if the eigenvector ψE is PT -symmetric, i.e. PT ψE = ψE . We speak
about unbroken PT -symmetry if all eigenvectors of H are PT -symmetric (and thus all
eigenvalues are real). Or equivalently, if every eigenvector of H is an eigenvector of
PT . Conversely, if this does not apply for any eigenvector of H we speak about broken
PT -symmetry.
2.2 Pseudo-Hermiticity
It is evident from above that the PT -symmetry is not a sufficient condition for the
reality of the spectrum since the eigenvectors do not have to be PT -symmetric. This is
also supported by known PT -symmetric models with a complex spectrum (e.g. previ-
ously mentioned model from [5] for ε < 0 ). Thus the search of properties of an operator
which would ensure the reality of the spectrum began. The main interest was aroused by
pseudo-Hermitian operators - many of the studied PT -symmetric Hamiltonians showed
this feature.
Definition 2.2.1 ([25]). Let T be a densely defined linear operator in H . T is called
pseudo-Hermitian (or η-pseudo-Hermitian), if there exists an operator η with properties
i) η, η−1 ∈ B(H ),
ii) T = η−1T ∗η,
iii) η = η∗.
This concept came from the physical background and over time it became clear that
there is an equivalent term in the mathematical literature - the J-self-adjointness in Krein
spaces. Krein space K is a complex linear space endowed with a self-adjoint sesquilinear
form [· , · ] (the so called indefinite inner product or indefinite metric). The positively
definite scalar product (· , · ) can be defined and the fundamental symmetry J having
properties J = J∗, J2 = I can be used to ensure the relation [φ, ψ] = (Jφ, ψ), φ, ψ ∈ K ,
between the former and the latter scalar product. Conversely, one can start with a
Hilbert space with a positively definite scalar product and fundamental symmetry and
define a Krein space.
10
The notion of an adjoint operator in Krein space is slightly different from the one in
Hilbert space since it refers to the indefinite inner product. The Krein space adjoint A[∗]
of a densely defined linear operator A in K is defined by
[Aφ,ψ] = [φ,A[∗]ψ] for φ ∈ D(A), ψ ∈ D(A[∗]) (2.5)
where D(A[∗]) := {ψ ∈ K |[A· , ψ] is continuous on D(A)}. The operator A is then
called self-adjoint in Krein space if A = A[∗]. It follows from this property that A∗ =
JAJ therefore this operator can be referred to as J-self-adjoint with respect to the
fundamental symmetry J . For more details about the relation of the PT -symmetry and
the Krein spaces, the reader is referred to [24] and references therein.
To illustrate the relation between the pseudo-Hermiticity in Hilbert space and the
J-self-adjointness in Krein space one can consider a η-pseudo-Hermitian operator A (i.e.
ηA = A∗η). Then we define Krein space using the self-adjoint form [x, y]η = (x, ηy) and
the fundamental symmetry J = η√η2−1
. It can be easily verified that J satisfies all the
requirements that it could be a conjugation operator in the Krein space. After further
algebraic manipulation it can be verified that JAJ = J [∗] and thus A is J-self-adjoint
in this space [3]. In the scope of PT -symmetric quantum mechanics is usually used the
operator P, it is therefore P-self-adjointness.
A special class of pseudo-Hermitian operators are Quasi-Hermitian operators. Their
importance was emphasized in [30] and since that time they hold an important place in
PT -symmetric quantum mechanics.
Definition 2.2.2 ([9]). Let T be a densely defined linear operator in H . T is called
quasi-Hermitian, if there exists an operator Θ with properties
i) Θ,Θ−1 ∈ B(H ),
ii) Θ ≥ 0,
iii) T = Θ−1T ∗Θ.The operator Θ is called the metric operator or just the metric.
The main reason why the quasi-Hermitian operators are relevant to our purpose is the
following: It was shown in [30] that for each irreducible set of quasi-Hermitean operator
there exists a unique metric operator. With this operator it is possible to modify the
scalar product
(· , · )Θ := (· ,Θ· ). (2.6)
11
In [30] it has been also proven the space with this metric is a Hilbert space. This enables
a different approach to quantum mechanics - the suitable set of observables (non-self-
adjoint but with real spectrum), for which a unique scalar product exists, is chosen, these
observables are self-adjoint with respect to this scalar product and the calculations are
then carried out in the Hilbert space defined by this scalar product.
The uniqueness is an important condition with regard to the statistical interpretation
of quantum mechanics - it is impossible for a consistent quantum theory to yield multi-
valent predictions. In other words, the expectation value of a physical observable could
be different with respect to each non-unique scalar product. Another point of view is to
regard the quasi-Hermitean operators as operators similar to the self-adjoint ones.
Theorem 2.2.3 ([4, Prop. 1]). Let T be a quasi-Hermitian operator with metric operator
Θ. Then T is similar to the self-adjoint operator H, T = ρHρ−1, where ρ =√T .
However, there still remains an open question - How to find the metric operator
Θ? A partial answer for a finite-dimensional spaces can be found in [26, 25] and it is
supplemented in [31]: Operator T is quasi-Hermitian if and only if its spectrum is real
and it is diagonalizable (i.e. its eigenvectors form an bi-orthonormal basis). The operator
Θ then has the form Θ =∑n
j=0 cj(φj , · )φj , where n is dimension of the Hilbert space, cj
are real positive constants and φj are eigenvectors of T ∗. For infinite-dimensional spaces
the general prescription for the metric operator exists only in special cases [21, 33, 28, 29]
and most of the known examples of the metric are just approximative and expressed as
the leading term of the perturbation series, e.g. in [6, 27].
2.3 C-self-adjointness
The substantial part of Hamiltonians studied in PT -symmetric examples satisfies
the relation H∗ = T HT . It is not necessarily limited to PT -symmetric models. This
property found its counterpart in mathematic literature in the term of C-self-adjointness
(not to be confused with the similar property associated with Krein spaces). The PT -
symmetry then can be understood as special case of C-self-adjointness.
Definition 2.3.1 ([10]). An operator C defined on Hilbert space H is a conjugation
operator, if for all φ, ψ ∈H
(Cφ,Cψ) = (ψ, φ),
C2φ = φ.
12
Definition 2.3.2 ([10]). A densely defined operator T in H is said to be C-symmetric
if CTC ⊂ T ∗. T is complex-symmetric if it is C-symmetric for a conjugation operator
C. T is said to be C-self-adjoint if CTC = T ∗.
The usefulness of this property is supported by the theorem proven in [7], which claims
that the residual spectrum of these operators is empty. This significantly facilitates
the spectral study of these non-self-adjoint operators. For more facts about complex
symmetric operator the reader is referred to [11, 13, 14]
13
Chapter 3
Physical realisation
Let us consider a charged quantum particle of mass m and electric charge e. We will
be interested in its interaction with a homogenous time-independent magnetic field. The
Hamiltonian for a classical particle in a general electromagnetic field is given by
H =1
2m(~p− e ~A(~x))2 + eφ(~x), (3.1)
where ~A(~x) is a vector potential and φ(~x) is a scalar potential. Their relation to the
vector of electric field intensity and vector of magnetic induction is ~E = −gradφ− 1c∂ ~A∂t
for the former and ~B = rot ~A for the latter. Using the correspondence principle, we can
find the corresponding quantum Hamiltonian
H =1
2m(~P − e ~A( ~X))2 + eφ( ~X)
=~P 2
2m− e
m( ~A( ~X). ~P ) +
i~e2m
div ~A( ~X) +e2
2m| ~A( ~X)|2 + eφ( ~X).
(3.2)
Capital ~P and ~X now denote the operators of momentum and position respectively,
Pjψ(~x) = −i~∂
∂xjψ(~x), Xjψ(~x) = xjψ(~x). (3.3)
The index j assumes values 1, 2, 3. If we now restrict ourselves to the case of homogenous
time-independent magnetic field, we can choose potentials due to their gauge invariance
so that φ( ~X) = 0 and ~A( ~X) = 12~B × ~X. Hamiltonian (3.2) then takes form
H =~P 2
2m− µ0
~B.~L+e2
8m( ~B × ~X)2, (3.4)
where ~L is the vector of the angular momentum of the particle (~L = ~X× ~P ) and µ0 is the
magneton of the particle (µ0 = e~2m). The last term of this expression can be neglected
for the usual values of magnetic induction (for more details on this step see [12]). We
shall do this.
14
Hamiltonian (3.4) was in the dawn of the quantum theory experimentally verified on
the Hydrogen atom. The theory claimed that the energy levels are going to split when the
particle will be inserted into the magnetic field. This phenomenon was later named the
Zeeman effect. In spite of the fact that the energy levels indeed split as it was predicted,
the number of the new energy levels did not correspond with the predicted number
and the energy of the ground state split as well as the other energy levels although it
was not supposed to do so [18]. The corrections in the theory naturally followed with
considerable support from the results of the Stern-Gerlach experiment. Another term
depending on a discovered new internal property of particles, the spin, was added to the
Hamiltonian.
Spin has nothing to do with the motion about the center of mass as it is in the
macroworld. Characteristics of this intrinsic angular momentum were derived by means
of the algebraic theory of spin in analogy with the algebraic theory of orbital angular
momentum ~L [16]. Each type of particles has its own specific value of the spin. Thanks
to these new degrees of freedom the description of the particle with a single wave function
is not complete and therefore it is necessary to use functions with several independent
components. Thus the operators need to be adjusted too - they become operator matri-
ces. The operators not related to the magnetic field are for this purpose a multiplication
of the identity matrix.
In the following text we are going to be concerned with particles with spin equal to 12 .
Let us choose an electron as their representative. In the case of spin 12 the wavefunctions
(called also spinors) have two components and the operators are 2 × 2 matrices. The
operator of spin ~S has three components
Sj =~2σj ,
where σj are the Pauli spin matrices
σ1 =
(0 1
1 0
), σ2 =
(0 −i
i 0
), σ3 =
(1 0
0 1
). (3.5)
The modified Hamiltonian (3.4) now reads
H =~P 2
2m− µ0
~B.~L− 2µ0~B.~S. (3.6)
The added multiplicative constant 2 may look a little bit odd but there are reasons for
adding it coming from quantum field theory. Reader interested in these reasons can find
them e.g. in [15].
15
a0−a
~B
x
Figure 3.1: Electron passing through a magnetic field ~B.
Assume that the electron flies into the homogenous magnetic field whose induction ~B
is oriented in the direction of motion. We presume that the wavefunction is separable
on three parts, each one depending only on one coordinate. We set the third coordinate
axis in the direction of motion. We assume that the magnetic field is supported in the
layer R2 × (−a, a). In this way, Hamiltonian for this direction will be diagonal, as the
first term in Hamiltonian (3.6) is a multiplication of the identity operator, the second
term disappears and the third one is composed of σ3. We therefore arrive at
H =
(− ~2
2m∆ + µ0|B| 0
0 − ~22m∆− µ0|B|
)(3.7)
defined on the Hilbert space L2((−a, a)). We shall be interested in this thesis only in
this one-dimensional scattering problem.
The Schrodinger equation for the spinor Ψ =(ψ+
ψ−
)reads
− ~2
2mψ′′+ + µ0|B|ψ+ = k2ψ+
− ~2
2mψ′′− − µ0|B|ψ− = k2ψ−,
where k is a positive wave number. Thanks to vanishing of the potential outside of the
interval (−a, a) we have the asymptotic solutions
Ψi =
(c1eikx
c2eikx
)+
(c3e−ikx
c4e−ikx
)(3.8)
for the in-coming wave and
Ψo =
(c5eikx
c6eikx
)(3.9)
16
for the out-coming wave. We introduced the complex constants cj , j = 1, 2, 3, 4, 5, 6,
which satisfy relations
|c1|2 + |c2|2 = 1,
|c3|2 + |c4|2 + |c5|2 + |c6|2 = 1.
The constant 1 determines just the normalization of the wavefunctions. This arrange-
ment allows us to send in wave packets only with one component of the spinor and
to study cases when the out-coming packet has only one component. We set further
restriction by demanding the perfect transmission (i.e. c3 = c4 = 0).
If we employ this in the in-coming wave and request continuity of Ψ and Ψ′ at the
boundary ±a, we reach the following non-linear problem− ~2
2mΨ′′ +
µ0|B| 0
0 −µ0|B|
Ψ = k2Ψ, in (−a, a),
Ψ′(±a)− ikΨ(±a) = 0.
(3.10)
This non-linear problem can be studied via a linear spectral problem− ~2
2mΨ′′ +
µ0|B| 0
0 −µ0|B|
Ψ = µ(α)Ψ, in (−a, a),
Ψ′(±a)− iαΨ(±a) = 0.
(3.11)
In this equation α is a real parameter and µ an eigenvalue depending on this parameter.
We can then find energies of the perfect transmission states as µ(α∗) for the points α∗
which satisfy
µ(α∗) = α2∗. (3.12)
In this manner we transformed the scattering problem at the interval (−∞,+∞) to
the spectral problem at the interval (−a, a) with Robin boundary conditions. This type
of boundary conditions has been appearing frequently in the course of the study of non-
self-adjoint operators (i.e. [21, 20, 22, 7, 8]). These boundary conditions convert to
Neumann boundary conditions for α tending to zero and to Dirichlet boundary condi-
tions for α tending to infinity. In this thesis we will nevertheless consider more general
cases of boundary conditions (see (5.2)) not necessarily with direct physical application.
Owing to these boundary conditions the probability flow at the boundary points ±adoes not disappear and therefore the non-self-adjointness can be interpreted as loss or
gain of probability density at the boundary points. More details about the transforming
scattering problems into PT -symmetric spectral problems can be found in [17].
17
Chapter 4
Elements of the theory of
sectorial forms
In the following paragraphs we will briefly summarize few basic properties of sectorial
forms and Sobolev spaces which will be used in the subsequent chapters. We denote by
H an arbitrary Hilbert space. Many of the following definitions and theorems can be
further generalized, however, the presented forms are sufficient for our purposes. Let us
start with the notion of a sesquilinear form.
Definition 4.0.1. A complex valued function t(φ, ψ) defined for φ, ψ ∈ D(t) ⊂ H is
called a sesquilinear form on H if it is antilinear in φ and linear in ψ. The function
t[φ] := t(φ, φ) is called a quadratic form.
The use of quadratic forms is convenient when we deal with Schrodinger operators
(such as the Hamiltonian (3.7)) since they require less regularity of functions in its
domain. Many differential operators with substantially different domains have quadratic
forms with the same domain. In comparison with a similar problem for operators, it is
not difficult to find the adjoint form, it is given by t∗(φ, ψ) := t(ψ, φ),D(t∗) := D(t). A
form t is said to be symmetric if t(φ, ψ) = t∗(φ, ψ) for all φ, ψ ∈ D(t) = D(t∗). Knowing
the adjoint form we can introduce the forms Re t := t+t∗
2 and Im t := t−t∗2i called the
real and the imaginary part of t, respectively. Let us note that neither Re t(φ, ψ) nor
Im t(φ, ψ) are real-valued and thus have nothing to do with Re (t(φ, ψ)) and Im (t(φ, ψ)).
They only satisfy the relations Re t[φ] = Re (t[φ]) and Im t[φ] = Im (t[φ]), which justify
the notation.
We will further introduce the notion of numerical range which is important in various
applications related to operators and forms in a Hilbert space.
18
Re η
Im η
Figure 4.1: The numerical range Θ(t) of a sectorial form t is enclosed in a sector.
Definition 4.0.2. Let t be a sesquilinear form in H . We call the set {t[φ] |φ ∈ D(t), ‖φ‖ = 1}the numerical range of t and denote Θ(t). Let T be an operator in H . We call the set
{(φ, Tφ) |φ ∈ D(T ), ‖φ‖ = 1} the numerical range of T and denote Θ(T ).
The numerical range need not to be open or closed, generally we can only say that
it is a convex set in the complex plane.
4.1 Sectorial forms
For the symmetric forms, it is quite easy to establish in a very natural way the concept
of boundedness from below since the quadratic form is real-valued. Our interest in forms
bounded from below arises from the physical motivation to the problem studied in this
thesis - the requirement of the ground state for the Hamiltonians considered in quantum
mechanics is very customary and it has to be also fulfilled by their corresponding forms.
In order to generalize this concept to the case of a nonsymmetric form we introduce the
sectorial forms.
Definition 4.1.1 ([19]). Let t be a sesquilinear form in the Hilbert space H . The form
t is said to be sectorially bounded from the left (or sectorial) if Θ(t) is a subset of a
sector {η ∈ C∣∣|arg(η − γ)| ≤ θ, 0 ≤ θ < π
2 , γ ∈ R} (see Figure 4.1), γ is called a vertex
of t and θ a semi-angle.
It follows from the definition that Re t ≥ γ and |Im t[φ]| ≤ (Re t[φ] − γ) for φ ∈ D(t).
In other words, the real part of the form t is bounded from below and the numerical
19
range is enclosed in a sector (see Figure 4.1). The numbers γ and θ are not uniquely
determined by t. It is easy to see that the reduction of θ can be compensated by the
reduction of γ.
To examine the sectoriality of the form using the perturbation theory we have to
introduce the notion of relative boundedness which specifies the relation between two
forms.
Definition 4.1.2 ([19]). Let t be a sectorial form in H . A form t′ in H is said to be
relatively bounded with respect to t (or t-bounded), if D(t′) ⊃ D(t) and
|t′[u]| ≤ a ‖u‖2 + b |t[u]| (4.1)
where u ∈ D(t) and a,b are nonnegative constants.
We make the use of this property in the following theorem which gives us few properties
of sums of forms.
Theorem 4.1.3 ([19, Thm. VI-1.33]). Let t be a sectorial form in H and let t′ be
t-bounded with b < 1 in (4.1). Then t+ t′ is sectorial. t+ t′ is closed if and only if t is
closed.
In the light of this theorem we can divide the examined form into two parts and then
consider the part t′ just as a perturbation of the form t.
To introduce a term similar to the sectoriality of a form is not so simple for operators.
In the following paragraph, we will step by step approach the notion of an m-sectorial
operator.
Definition 4.1.4 ([10]). A linear operator T acting in a Hilbert space H is said to be
accretive if Re (ψ, Tφ) ≥ 0 for all φ ∈ D(T ), and quasi-accretive if T + αI is accretive
for some α > 0.
Definition 4.1.5 ([10]). A linear operator T ∈ C (H ) is said to be m-accretive if it
satisfies
{λ |Reλ < 0} ⊂ ρ(T ),
‖(T − λI)−1‖ ≤ 1
|Reλ| for Reλ < 0.
If T + αI is m-accretive for some α > 0, then T is said to be quasi-m-accretive.
20
Every m-accretive operator is accretive, hence its designation is justified. The notions
of accretive and m-accretive operators are often used in problems concerned with the
solvability of first- and second-order evolution equations and with differential equations
of elliptic type. Hence these types of operators were deeply studied, many of their
properties are known, but let us just state that if T is accretive then T ∗ is also accretive.
Definition 4.1.6 ([10]). A linear operator T in a Hilbert space H is said to be sectorial if
its numerical range lies in a sector {η ∈ C∣∣Re η ≥ γ, |arg(η − γ)| ≤ θ, 0 ≤ θ < π
2 , γ ∈ R}.We say that T is m-sectorial if it is sectorial and quasi-m-accretive.
The m-sectorial operators add to accretive operators further restrictions on the numerical
range. They are often generated by second-order differential expressions with some
boundary conditions. An important property of an m-sectorial operator is that it is
closed.
It is known for a bounded form t that there exists a bounded operator T such that
t(φ, ψ) = (φ, Tψ). This theorem can be generalized to a densely defined, sectorial and
closed form. The corresponding operator turns out to be not only closed and sectorial,
as it would be expected, but even m-sectorial.
Theorem 4.1.7 ([19, Thm. VI-2.1], The first representation theorem). Let t be a densely
defined, closed, sectorial sesquilinear form in H . There exist an m-sectorial operator T
such that
i) D(T ) ⊂ D(t) and t(u, v) = (u, Tv) for every u ∈ D(t) and v ∈ D(T );
ii) D(T ) is a core of t;
iii) if v ∈ D(t), w ∈ H and t(u, v) = (u,w) holds for every u belonging to a core of t,
then v ∈ D(T ) and Tv = w.
The m-sectorial operator T is uniquely determined by the condition i).
An essential tool for investigating the spectrum of a closed operator is its resolvent.
It captures the spectral properties of an operator in its analytic structure. Among
other applications not related to this thesis belongs solving the inhomogenous Fredholm
integral equations and finding the spectral decomposition of an operator.
Definition 4.1.8. Let T be a closed operator. The set ρ(T ) := {λ ∈ C|(T − λ)−1 exists
and is bounded } is called the resolvent set of T . The function RT defined on ρ(T ) by
RT (λ) := (T − λ)−1 is called the resolvent of T .
21
The spectrum of on operator T is the set σ(T ) := C \ ρ(T ). The spectrum is always
a closed set. It can be further separated into three parts - the point spectrum σp(T )
of such points λ (called also eigenvalues) that T − λ is not injective, the continuous
spectrum σc(T ) of such points λ that T − λ is not surjective and Ran(T ) is dense in
H and the residual spectrum σr(T ) of such points λ that T − λ is not surjective and
Ran(T ) is not dense in H . All these sets are disjoint.
The investigation of properties of the resolvent RT can lead to valuable information
about the spectrum of T . The knowledge that the resolvent is compact for some λ turns
out to be very useful. Then the resolvent is compact for all λ ∈ ρ(T ) and the spectrum of
the operator T consists entirely of isolated eigenvalues with finite algebraic multiplicities.
We can use the following theorem to verify the compactness of the resolvent.
Theorem 4.1.9 ([19, Thm. VI-3.4]). Let s be a densely defined, closed sectorial form
with Re s ≥ 0 and let S be the associated m-sectorial operator. Let t′ be a t-bounded
form satisfying (4.1) with b < 12 . Then t = s + t′ is also sectorial and closed. Let T
be the associated m-sectorial operator. Then the resolvents RS and RT exists. If S has
compact resolvent, the same is true of T .
4.2 Sobolev spaces
Sobolev spaces are vector space whose elements are complex functions defined on domains
in Rn. Their partial derivatives are required to satisfy certain integrability conditions.
The importance of Sobolev spaces lies in the fact that sometimes it is not sufficient to
deal with the classical solutions of differential equations and it is necessary to work with
distributions. We avoid in this text the general theory of Sobolev spaces and proofs of
theorems and state just the notions relevant to this thesis. Interested reader can find
more details on this theory in [1].
Let us start with the Hilbert space H := L2((−a, a)) with the norm ‖· ‖. We call
distribution every linear functional projecting smooth functions with a compact support
contained in (−a, a) into complex numbers. Every function f defined on (−a, a) can be
regarded as a distribution φf determined by the formula
φf (g) :=
∫ a
−af(x)g(x) dx,
22
where g is a smooth function with a compact support in (−a, a). The weak derivative
Dαφ for arbitrary α ∈ N and distribution φ is defined by
(Dαφ)(g) := (−1)αφ(Dαg).
Let us now introduce the Sobolev space Hn. It is a special case of the more general
Sobolev space Wn,p for p = 2. We define the space
Hn((−a, a)) := {f ∈ L2((−a, a))∣∣Dαf ∈ L2((−a, a)), 0 ≤ α ≤ n}.
The norm is defined by the means of the formula
‖f‖n,2 :=
∑α≤n‖Dαf‖2
12
. (4.2)
It can be proven that H2((−a, a)) is a separable Hilbert space. It is obvious that it
is imbedded in L2((−a, a)) and it holds it is a dense subset. Due to one of the main
results of the theory, the Sobolev imbedding theorem [1, Thm. 4.12], it holds that Hn
is imbedded in the space consisting of functions having bounded, uniformly continuous
derivatives up to order n− 1 on [−a, a].
23
Chapter 5
Definition of the Hamiltonian
In this chapter we introduce the proper definition of the Hamiltonian (3.7) which arose
from the physical motivation to the problem and we establish its basic properties. This
Hamiltonian acts in the Hilbert space H := L2((−a, a)) ⊗ C2 ' L2((−a, a);C2) on
spinors Ψ =(ψ+
ψ−
). Scalar product in this space is defined as
(Φ,Ψ) :=
∫ a
−aΦT
(x)Ψ(x) dx. (5.1)
The corresponding norm is given as usual by ‖Ψ‖ =√
(Ψ,Ψ). The absolute value of
spinor will be understood as |Ψ(x)| =√|ψ+(x)|2 + |ψ−(x)|2 for x ∈ [−a, a]. We impose
Robin-type boundary conditions (RTBC),
Ψ′(a) +AΨ(a) = 0,
Ψ′(−a)−AΨ(−a) = 0,(5.2)
where A ∈ C2×2,
A =
(p q
r s
), p, q, r, s ∈ C. (5.3)
This leads us to the definition of H
HΨ :=
(−ψ′′+ + cψ+
−ψ′′− − cψ−
),
D(H) :=
{Ψ ∈ H2((−a, a);C2)
∣∣∣∣∣ Ψ′(a) +AΨ(a) = 0
Ψ′(−a)−AΨ(−a) = 0
},
(5.4)
where c := µ0|B| = e~2m |B|. In order to simplify the form of the Hamiltonian, we used
units in which ~ = 1 and m = 1/2. We will keep this units for the rest of this thesis.
24
5.1 Definition via quadratic form
Equipped with the Hamiltonian, we can take a look how its associated sesquilinear form
looks like. Starting from a general prescription for sesquilinear form h(Φ,Ψ) we obtain
for all Ψ ∈ D(H) and Φ ∈ H1((−a, a);C2)
(Φ, HΨ) =
∫ a
−aΦT
(x)HΨ(x) dx
=
∫ a
−a
(−φ+(x)ψ′′+(x) + cφ+(x)ψ+(x)− φ−(x)ψ′′−(x)− cφ−(x)ψ−(x)
)dx
=c(φ+, ψ+)− c(φ−, ψ−)−∫ a
−a
(φ+(x)ψ′′+(x) + φ−(x)ψ′′−(x)
)dx
=c(φ+, ψ+)− c(φ−, ψ−)− [φ+(x)ψ′+(x)]a−a − [φ−(x)ψ′−(x)]a−a
+
∫ a
−a
(φ′+(x)ψ′+(x) + φ′−(x)ψ′−(x)
)dx
=(Φ′,Ψ′) + c(φ+, ψ+)− c(φ−, ψ−)− ΦT
(a)Ψ′(a) + ΦT
(−a)Ψ′(−a).
If we insert the boundary condition (5.2) into this equation, we get
(Φ, HΨ) = (Φ′,Ψ′) + c(φ+, ψ+)− c(φ−, ψ−) + ΦT
(a)AΨ(a) + ΦT
(−a)AΨ(−a). (5.5)
Now we can move to the quadratic form by replacing Φ by Ψ in equation (5.5)
(Ψ, HΨ) =‖Ψ′‖2 + c‖ψ+‖2 − c‖ψ−‖2 + ΨT
(a)AΨ(a) + ΨT
(−a)AΨ(−a)
=‖Ψ′‖2 + c‖ψ+‖2 − c‖ψ−‖2
+ p|ψ+(a)|2 + qψ+(a)ψ−(a) + rψ−(a)ψ+(a) + s|ψ−(a)|2
+ p|ψ+(−a)|2 + qψ+(−a)ψ−(−a) + rψ−(−a)ψ+(−a) + s|ψ−(−a)|2.
(5.6)
In order to see properties of the Hamiltonian (5.4) let us now start from the other
end. We define sesquilinear form by the right-hand side of (5.5)
h(Φ,Ψ) := (Φ′,Ψ′) + c(φ+, ψ+)− c(φ−, ψ−)− ΦT
(a)Ψ′(a) + ΦT
(−a)Ψ′(−a),
D(h) := H1((−a, a);C2),(5.7)
without a priori knowing the operator from which it emerged. Notice that the boundary
terms are well defined thanks to embedding of H1((−a, a);C2) in the space of uniformly
continuous functions on [−a, a] (see Chapter 4 for more details). It should be stressed out
that form h has weaker requirements on the regularity of the functions than Hamiltonian
H. Our first main goal is to show that this form is sectorial. To do this we prove the
following auxiliary claim.
25
Lemma 5.1.1. The inequality |Ψ(±a)|2 ≤ 12a‖Ψ‖2 + 2‖Ψ‖‖Ψ′‖ holds for all Ψ ∈ D(h).
Proof. We make the following proof for x = a. The case of point −a can be proven with
a similar course of steps. We choose the auxiliary function η so that it complies
η(x) =
0 x ∈ (−∞,−a)
x+a2a x ∈ [−a, a)
1 x ∈ [a,∞)
Then it is possible to make the following estimate for Ψ(a)
|Ψ(a)|2 =
∫ a
−a
d
dx
(η(x)|Ψ(x)|2
)dx
=
∫ a
−aη′(x)|Ψ(x)|2 dx+ 2
∫ a
−aη(x)|Ψ(x)||Ψ′(x)|dx
≤ 1
2a‖Ψ‖2 + 2
∫ a
−a|Ψ(x)||Ψ′(x)| dx
≤ 1
2a‖Ψ‖2 + 2‖Ψ‖‖Ψ′‖.
The Schwarz inequality∫ a−a |Φ(x)||Ψ(x)| dx ≤
(∫ a−a |Φ(x)|2 dx
) 12(∫ a−a |Ψ(x)|2 dx
) 12
which
holds for every Φ,Ψ ∈H was used in the last step.
Let us now divide the form (5.7) into three parts:
h1[Ψ] = ‖Ψ′‖2
h2[Ψ] = i Im(
ΨT
(a)AΨ(a) + ΨT
(a)AΨ(−a))
h3[Ψ] = c‖ψ+‖2 − c‖ψ−‖2 + Re(
ΨT
(a)AΨ(a) + ΨT
(−a)AΨ(−a)).
(5.8)
This division is going to be used in the forthcoming theorems. Note that h1 corresponds
to Neumann Laplacian [10] and that h2[Ψ] is purely imaginary whereas h1[Ψ] and h2[Ψ]
are real (see Chapter 4 for the definition of real and imaginary part of a form).
Theorem 5.1.2. The form h defined by (5.7) is densely defined, closed and sectorial.
Proof. The domain of h, H1((−a, a);C2), is dense in H , hence h is densely defined.
Using Lemma 5.1.1 we can carry out an upper bound on h2 and h3
|h2[Ψ]| ≤ |A|(|Ψ(a)|2 + |Ψ(−a)|2) ≤ |A|a‖Ψ‖2 + 4|A|‖Ψ‖
√h1[Ψ],
|h3[Ψ]| ≤ c‖Ψ‖2 +|A|a‖Ψ‖2 + 4|A|‖Ψ‖
√h1[Ψ],
26
Reh[Ψ]
Imh[Ψ]
Figure 5.1: The numerical range Θ(h) of the sectorial form h is enclosed in a parabola.
It can be further restricted with the use of Lemma 5.1.4.
where |A| := |p|+ |q|+ |r|+ |s|. Putting these two estimates together we obtain
|(h2 + h3)[Ψ]| ≤ 8|A|‖Ψ‖√h1[Ψ] +
(2|a|a
+ c
)‖Ψ‖2
≤ 4|A|εh1[Ψ] +
(4|A|ε
+2|a|a
+ c
).
In the last step we used the Young inequality 2ab ≤ εa2+ b2
ε which holds for all a and b real
and ε > 0 for a =√h1[Ψ] and b = ‖Ψ‖. This gives us that h2+h3 is h1-bounded with the
relative bound arbitrarily small since ε can be chosen arbitrarily small. h1 corresponds
to Neumann Laplacian in H and thus is densely defined, symmetric, positive and closed
[10, Section IV.]. It follows easily from this that it is also sectorial. Therefore the form
h = h1 + h2 + h3 is according to Theorem 4.1.3 closed and sectorial.
From Chapter 4 we know that the numerical range of a sectorial form is in a sector.
However, in this case we can improve the estimate of the numerical range - instead of
the sector we can enclose it in a parabola in the complex plane (see Figure 5.1). Thus
the numerical range has much better behavior at the infinity.
Theorem 5.1.3. The form (5.7) satisfies relation |Imh[Ψ]| ≤ C1
√Reh[Ψ] + C2 + C3,
where C1, C2, C3 > 0.
Proof. Because the imaginary part of h is in fact h2 we seek an upper bound containing
square root of h1 + h3. We recall the estimate of h2 and h3 which we obtained in the
27
proof of Theorem 5.1.2 considering we can normalize the function Ψ to 1
|h2[Ψ]| ≤ |A|a
+ 4|A|√h1[Ψ],
|h3[Ψ]| ≤ c‖Ψ‖2 +|A|a
+ 4|A|√h1[Ψ].
This estimate of h2 is not yet what we desire, the real part of h is also composed of h3.
If we realise that the square root is an increasing function, we see that it is sufficient to
proof
h1[Ψ] ≤ k1 (h1[Ψ] + h3[Ψ]) + k2
for some positive constants k1 and k2. To check this we carry on a lower estimate of
Reh
Reh[Ψ] = (h1 + h3)[Ψ] ≥ h1[Ψ]−(c+|A|a
+ 4|A|√h1[Ψ]
)≥ (1− ε)h1[Ψ]−
(c+|A|a
+4|A|2ε
)=
1
2h1[Ψ]−
(c+|A|a
+ 8|A|2).
In the last equality we set ε = 12 . This is exactly what we were trying to show. We can
easily express h1 from this equation
0 ≤ h1[Ψ] ≤ 2Reh[Ψ] + 2
(c+|A|a
+ 8|A|2). (5.9)
Altogether we obtained
|Imh[Ψ]| ≤√
8|A|√
Reh[Ψ] +
(c+|A|a
+ 8|A|2)
+|A|a. (5.10)
Note that the square root is well defined thanks to the estimate (5.9).
We also know from Lemma 5.1.4 that the numerical range is symmetric with respect to
the real axis (the PT -symmetry of the Hamiltonian is going to be presented in Chapter
6). Because the parabola is shifted, the numerical range is further restricted (see Figure
5.1).
Lemma 5.1.4. Let h(φ, ψ) = (φ,Hψ) for all φ ∈ D(h), ψ ∈ D(H), where H is PT -
symmetric. Then Θ(h) is symmetric with respect to the real axis.
Proof. We recall from Chapter 2 that a PT -symmetric operator satisfies [H,PT ] = 0
and that (PT )2 = I. The numerical range Θ(h) is from definition given as the set of
28
h[ψ] = (ψ,Hψ), where ‖ψ‖ = 1. Let us begin with such ψ. Then we can look at the
point of Θ(h) associated with this vector
h[ψ] = (ψ,Hψ) = (PT PT ψ,PT PT HPT PT ψ) = (PT ψ,PT Hψ)
= (T ψ, T Hψ) = (Hψ, ψ) = (ψ,Hψ),
where we denoted ψ := PT ψ. Our goal is to prove that Reh[ψ] − iImh[ψ] also lies in
Θ(h). This is indeed satisfied when we use ψ
Reh[ψ] =(ψ,Hψ) + (ψ,Hψ)
2=
(ψ,Hψ) + (ψ,Hψ)
2= Reh[ψ]
Imh[ψ] =(ψ,Hψ)− (ψ,Hψ)
2i=
(ψ,Hψ)− (ψ,Hψ)
2i= −Imh[ψ]
We will show that sesquilinear form (5.7) defines the Hamiltonian (5.4). Using Theo-
rem 4.1.7 we know that there exists an m-sectorial operator T defined via this form
TΨ := η,
D(T ) := {Ψ ∈ D(h) | ∃ζ ∈H ,∀Φ ∈ D(h), h(Φ,Ψ) = (Φ, ζ)}.(5.11)
By definition, T further satisfies the relation h(Φ,Ψ) = (Φ, TΨ) for all Φ ∈ D(h) and
Ψ ∈ D(T ).
Theorem 5.1.5. The Hamiltonian H defined by (5.4) equals to the operator T defined
by (5.11).
Proof. By integration by parts it can be directly shown that H ⊂ T . This is essentially
the inverse process to that from which we originally derived the sesquilinear form.
It remains to prove T ⊂ H. The course of this proof is inspired by [19, Example
VI.2.16]. In order to do so we come out of the Relation h(Φ,Ψ) = (Φ, ζ) which means∫ a
−aΦTζ dx =
∫ a
−a
(Φ′T
Ψ′ + ΦTCΨ)
dx+ ΦT
(a)AΨ(a) + ΦT
(−a)AΨ(−a) (5.12)
where Φ =(φ+φ−
)∈ D(h) and Ψ =
(ψ+
ψ−
)∈ D(T ) are functions of x and C =
(c 00 −c
)is
a constant matrix. Let z(x) :=∫ x−a (ζ − CΨ) dy. Then∫ a
−aΦT
(ζ − CΨ) dx =
∫ a
−aΦTz′dx
=[ΦTz]a−a−∫ a
−aΦ′Tz
= ΦT
(a)z(a)− ΦT
(−a)z(−a)−∫ a
−aΦ′Tz dx.
(5.13)
29
Putting equations (5.12) and (5.13) together yields∫ a
−aΦ′T (z + Ψ′
)dx+ Φ
T(a) (AΨ(a)− z(a)) + Φ
T(−a)
(AΨ(−a) + z(−a)
)= 0 (5.14)
for every Φ ∈ D(h) and Ψ ∈ D(T ). For any Φ′ ∈ L2((−a, a);C2) such that Φ(x) =∫ x−a Φ′(x) dx,
∫ a−a Φ′(x) dx = 0 is valid, Φ lies in D(h) and satisfies Φ(a) = Φ(−a) = 0.
It follows from equation (5.14) that (z + Ψ′) ⊥ Φ′ for this special choice of Φ. Thus
z + Ψ′ ∈ {1⊥}⊥ = span{1} since
(1,Φ′) =
∫ a
−aΦ′(x) dx = 0.
In other words z+Ψ′ equals to a constant k. Putting this into equation (5.14) we obtain
by integration
ΦT
(a) (AΨ(a)− z(a) + k) + ΦT
(−a)(AΨ(−a) + z(−a)− k
)= 0.
Since Φ(a) and Φ(−a) can be any complex numbers when Φ varies over D(h), their
coefficients must vanish. Together with the fact that k = z(a)+Ψ′(a) = z(−a)+Ψ′(−a)
we get
Ψ′(a) +AΨ(a) = 0,
Ψ′(−a)−AΨ(−a) = 0.(5.15)
Further we can see that Ψ′ is absolutely continuous and Ψ′′ = −z′ = −ζ + CΨ ∈L2((−a, a),C2, dx)). So for every Ψ ∈ D(T ) it holds that Ψ and Ψ′ are absolutely
continuous and Ψ′′ ∈ L2((−a, a),C2, dx). It further satisfies the boundary conditions
(5.15). Ψ therefore belongs to D(H) and
TΨ = ζ = −Ψ′′ + CΨ.
These propertied directly show that T ⊂ H since D(T ) ⊂ D(H) and TΨ = HΨ for all
Ψ ∈ D(T ).
5.2 Adjoint operator
Equipped with the First representation theorem we can take a look on the adjoint
operator to Hamiltonian (5.4). Finding it can be in general difficult task but we opened
a new way for finding it with its proper definition by sesquilinear form in Theorem 5.1.5.
30
Theorem 5.2.1. The adjoint operator to Hamiltonian H defined by (5.4) is the operator
H∗ defined by
H∗Ψ =
(−ψ′′+ + cψ+
−ψ′′− − cψ−
),
D(H∗) =
{Ψ ∈ H2((−a, a);C2)
∣∣∣∣∣ Ψ′(a) +A∗Ψ(a) = 0
Ψ′(−a)−ATΨ(−a) = 0
},
(5.16)
Proof. We can find the adjoint form to the form h from the definition
h∗(Φ,Ψ) := h(Ψ,Φ) = (Φ′,Ψ′)+c(φ+, ψ+)−c(φ−, ψ−)+Ψ(a)TAΦ(a)+ΨT (−a)AΦ(−a),
where h(Φ,Ψ) is the form (5.7). Basic properties of scalar product were used. The last
two terms in this expression can be adjusted with basic matrix operations
Ψ(a)TAΦ(a) = ΦT
(a)A∗Ψ(a),
ΨT (−a)AΦ(−a) = ΦT
(−a)ATΨ(−a).
after this algebraic manipulation, we obtain expression very similar to the original form
(5.7). The only difference is the substitution of matrix A for matrix A∗. The expression
for the adjoint operator follows from Theorem 4.1.7 and the proof of Theorem 5.1.5.
With the knowledge of adjoint operator we can discuss when the Hamiltonian is D-
self-adjoint for different choices of operator D. It turns out that impose conditions for
the matrix A.
Proposition 5.2.2. The Hamiltonian H is self-adjoint if and only if and only if A = A∗.
Proof. We can directly compare both operators, we then see that they are identical when
A = A∗.
Proposition 5.2.3. The Hamiltonian H is P-self-adjoint if and only if A = AT .
Proof. Let us remind the relation satisfied by every P-self-adjoint operator:
H∗ = PHP.
It is not hard to verify that if Ψ ∈ H2((−a, a);C2) then also PΨ ∈ H2((−a, a);C2).
Since the second derivative and the constant c are symmetric with respect to the parity,
we have to check only the boundary conditions (5.2) and see how the vector PΨ fits in:
Ψ′(a) +AΨ(a) = 0,
Ψ′(−a)−AΨ(−a) = 0.
31
By comparing this result with the original boundary condition (5.2) we see when the
domains of H∗ and HP coincide and we get the condition on P-self-adjointnes A =
AT .
Proposition 5.2.4. The Hamiltonian H is T -self-adjoint if A = A.
Proof. The proof is analogous to the proof of Proposition 5.2.3
32
Chapter 6
Spectral analysis
We approach the spectral analysis of Hamiltonian (5.4). Since this thesis is interested in
operators having PT -symmetry, let us check that our Hamiltonian fulfills this property.
Lemma 6.0.5. The operator H defined by (5.4) is PT -symmetric.
Proof. We know from Definition 2.2 that H is PT -symmetric if [H,PT ] = 0. This
relation includes that for all Ψ ∈ D(H) holds that PT Ψ ∈ D(H). We denote Φ(x) :=
(PT Ψ)(x) = Ψ(−x). With regard to properties of Sobolev spaces as mentioned in
Chapter 4 we first must check that Φ,Φ′ and Φ′′ belong to L2((−a, a)). This is trivially
satisfied since it holds for Ψ. Φ should also satisfy RTBC (5.2) at ±a. We shall check
this for the point a
Φ′(a) +AΦ(a) = −Ψ′(−a) +AΨ(−a)
= −Ψ(−a) + Re (A)Ψ(−a) + i Im (A)Ψ(−a)
= −Ψ′(−a) + Re (A)Ψ(−a)− i Im (A)Ψ(−a) = 0.
The proof for the point −a is analogous. The statement now follows from the fact that
the second derivative and the constant c commute with the operator PT .
The spectrum turns out to be discrete as it would be expected since our PT -symmetric
Hamiltonian is a differential operator on the compact interval. Nevertheless, there are
few known examples of PT -symmetric models with continuous spectrum [2, 7, 23].
Theorem 6.0.6. The operator H defined by (5.4) is an operator with compact resolvent.
Proof. According to Theorem (5.1.5), the operator H corresponds to the form (5.7). We
recall the division of this form and the estimate carried out in the proof of the Theorem
(5.1.2) which gives us that h2+h3 is h1-bounded with the relative bound arbitrarily small.
33
It can be thus set smaller than 12 . Since h1 corresponds to the Neumann Laplacian
in H and thus is densely defined, closed and positive [10, Section IV.], it meets the
requirements of Theorem 4.1.9. Because the Neumann Laplacian is an operator with
compact resolvent, it follows from this theorem that H is an operator with compact
resolvent.
Corollary 6.0.7. The spectrum of H consists entirely of isolated eigenvalues with finite
algebraic multiplicities.
We will try to broaden our knowledge about the point spectrum and try find the explicit
form of the eigenvalues. Nevertheless, the latter is possible only in special cases of the
RTBC (5.2), in the general case it turns out to be impossible. In general, our only way
how to describe the eigenvalues is by means of an implicit equation.
Theorem 6.0.8. The eigenvalues of H defined by (5.4) are determined by the equation((pk+ − sk−)2 − (ps− qr + k+k−)2
)cos (2a (k− − k+))
−(
(pk+ + sk−)2 − (ps− qr − k+k−)2)
cos (2a (k− + k+))
+ 2 (pk+ − sk−) (ps− qr + k+k−) sin (2a (k− − k+))
− 2 (pk+ + sk−) (ps− qr − k+k−) sin (2a (k− + k+))
+ 4qrk+k− = 0.
(6.1)
The corresponding eigenfunctions are
Ψ =
(A+ cos(k+x) +B+ sin(k+x)
A− cos(k−x) +B− sin(k−x)
)(6.2)
where k+ =√λ+ c and k− =
√λ− c.
Proof. We are looking for the solution Ψ of the equation HΨ = λΨ in the form (6.2).
Substituting this expression into the RTBC (5.2) we get equation
M
A+
B+
A−
B−
= 0, (6.3)
where M ∈ C4×4 is the matrix taking the form(p+ ik−)eiak− (p− ik−)e−iak− qeiak+ qe−iak+
reiak− e−iak− (s+ ik+)eiak+ (s− ik+)e−iak+
(−p+ ik−)e−iak− (−p− ik−)eiak− −qe−iak+ −qeiak+
−re−iak− −reiak− (−s+ ik+)e−iak+ (−s− ik+)eiak+
. (6.4)
34
0 1 2 3 4 5Α
5
10
15
20
25
Λ
Figure 6.1: α-dependence of eigenvalues for c = 0.4, a = π4 in Example a).
Nontrivial solution exists if and only if det(M) = 0. This yields the equation (6.1). The
eigenvectors are found by solving equation (6.3) with A− (eventually with A− and A+)
as a parameter.
6.1 Examples of boundary conditions
In the following text we abandon the generality and examine few specific simple exam-
ples of RTBC (5.2) using Proposition 5.2.2 and Theorem 6.0.8. As the formulae simplify
significantly, we continue in using notation k+ =√λ+ c and k− =
√λ− c. Furthermore
α always stands for a real positive parameter.
a)
A =
(α 0
0 α
)In this case we know from Proposition 5.2.2 that the eigenvalues are real. Equation
(6.1) simplifies significantly and reads
(2αk+ cos(2ak+)− (k2+ − α2) sin(2ak+))
(2αk− cos(2ak−)− (k2− − α2) sin(2ak−)) = 0.
(6.5)
The dependence of these eigenvalues on parameter α can be seen in Figure 6.1.
Equation (6.5) is composed of the product of two terms. In order to satisfy the
equation, at least one term in one of the two brackets must be equal to zero. The
dashed line in the figure corresponds to the case when this condition is satisfied for
the former, the full line for the latter. Due to separating both parts of the spinor in
35
0 2 4 6 8Α
5
10
15
20
Λ
i)
1.90 1.95 2.00 2.05 2.10Α
3.9
4.0
4.1
4.2
Λ
ii)
Figure 6.2: In i) α-dependence of eigenvalues for c = 1, a = π4 in Example b). In ii) the
detail of an avoided crossing of the first pair of eigenvalues.
the boundary condition, the final form of corresponding eigenvector depends on two
parameters:
Ψ =
(A+ cos(k+x) + −α cos(ak+)+k+ sin(ak+)
k+ cos(ak+)+α sin(ak+) A+ sin(k+x)
A− cos(k−x) + −α cos(ak−)+k− sin(ak−)k− cos(ak−)+α sin(ak−) A− sin(k−x)
).
b)
A =
(0 iα
−iα 0
)As well as in the previous example, we know in this one that the spectrum is real
because the matrix defining boundary conditions is self-adjoint, cf. Proposition 5.2.2.
The implicit equation for the eigenvalues now takes form
2α2k+k−(1− cos(2ak+) cos(2ak−)) = −(k2+k
2− + α4) sin(2ak+) sin(2ak−). (6.6)
The dependence of these eigenvalues on parameter α can be seen in Figure 6.2. An
interesting phenomenon in this figure is an approaching of a pair of eigenvalues and
its subsequent moving back and slowly approaching to constant values. It should be
noted that in the point of closest approach the two curves do not intersect. This
avoided crossing holds for each pair of the eigenvalues. The exact form of the eigen-
vector for these eigenvalues depends solely on one parameter
Ψ =
A cos(k+x) + k−k+ sin(ak−) sin(ak+)+α2 cos(ak+) sin(ak−) tan(ak−)k−k+ cos(ak+) sin(ak−)+α2 cos(ak−) sin(ak+)
A sin(k+x)iα sin(ak+)(k−k+ sin(ak+)+α2 cos(ak+) tan(ak−))k−(k−k+ cos(ak+) sin(ak−)+α2 cos(ak−) sin(ak+))
A cos(k−x)− iα cos(ak+)k− cos(ak−)A sin(k−x)
.
36
0 1 2 3 4 5 6Α
5
10
15
20
25
Λ
Figure 6.3: α-dependence of eigenvalues for c = 0.4, a = π4 , β = 0 in Example c).
c)
A =
(iα+ β 0
0 iα+ β
)Although the Hamiltonian satisfying these boundary conditions is not self-adjoint,
we can still use the fact that they separate both parts of the spinor. Equation (6.1)
can be rewritten in this case as(−2βk− cos(2ak−) + (k2
− − α2 − β2) sin(2ak−))(
−2βk+ cos(2ak+) + (k2+ − α2 − β2) sin(2ak+)
)= 0.
(6.7)
As a matter of fact, an eigenvalue problem corresponding to a very similar equa-
tion has been previously studied in [21] and in more detail in [22]. After a slight
modification of results of [22] to our case the corresponding eigenfunctions read
Ψ =
(A+ cos(k+x) + k+ sin(ak+)−(iα+β) cos(ak+)
k+ cos(ak+)+(iα+β) sin(ak+)A+ sin(k+x)
A− cos(k−x) + k− sin(ak−)−(iα+β) cos(ak−)k− cos(ak−)+(iα+β) sin(ak−)A− sin(k−x)
).
In further text we still suppose α and β real parameters. As we shall see, the spectrum
has significantly different characteristics for different fixed values of parameter β.
The simplest case is β = 0. As it has been shown in [21] (for one dimensional case
without magnetic field) and as it is seen from Figure 6.3 - one eigenvalue depends
od parameter α quadratically and the others are constant. These eigenvalues can be
expressed explicitly as
λj,± =
α2 ∓ c,(jπ2a
)2∓ c.
(6.8)
37
0 1 2 3 4 5 6Α
5
10
15
20
25
Λ
Figure 6.4: α-dependence of eigenvalues for c = 0.4, a = π4 , β = 0.5 in Example c).
0 1 2 3 4 5 6Α
10
20
30
40
Re Λ
1 2 3 4 5 6Α
-6
-4
-2
2
4
6
Im Λ
Figure 6.5: α-dependence of eigenvalues for c = 0.4, a = π4 , β = −0.5 in Example c).
Because equation (6.7) is composed of two terms, we can as in Example a) distinguish
between cases when the first parenthesis is equal to zero and when this is true for the
second one. Coherently with Example a), the dashed line in the figure corresponds
with the case when the former case applies, the full line with the case when the latter
applies.
The reality of the spectrum was proved for the case β > 0 in [22]. The Figure 6.4
shows dependance of these eigenvalues on parameter α. We can again observe the
pairs of eigenvalues at a distance of 2c.
However, the reality of spectra in the case when β < 0 is not guaranteed and indeed
it is easily seen from the Figure 6.5 that complex conjugated pairs of eigenvalues
appear and they have non-trivial imaginary part for large range of values of parameter
α. This is manifested when two eigenvalues collide - then the imaginary part arises.
38
The situation when this pair of complex eigenvalues returns to real numbers can be
also observed. It should be also noted that only one pair of complex conjugated
eigenvalues occurs simultaneously in the spectrum [22].
This example of the boundary conditions can be easily interpreted physically in
the scattering interpretation. We recall from Chapter 3 the boundary conditions in
the equation (3.11), Ψ′(±a) − iαΨ(±a) = 0. We consider adding two Dirac delta-
interactions of strength β ∈ R located in the points −a, a. So the wavefunctions must
further satisfy
Ψ′(−a−)−Ψ′(−a+) = βΨ(−a),
Ψ′(a−)−Ψ′(a+) = βΨ(−a).
It follows from the continuity of Ψ and Ψ′ at the boundary ±a out more general
example of boundary conditions.
d)
A =
(0 iα
iα 0
)This boundary conditions quite resemble those of case b). The equation determining
the eigenvalues looks similar as well:
2α2k+k−(1− cos(2ak+) cos(2ak−)) = (k2+k
2− + α4) sin(2ak+) sin(2ak−). (6.9)
That would draw one on the conclusion that the eingenvalues will also be real as the
left-hand side of the equation contains trigonometric functions identical to those of
the case b) and the right-hand side changed only the sign. This assumption would
be wrong, since numerical simulations suggest that there appear complex eigenvalues
(see Figure 6.6). The eigenvalues exhibit interesting feature - unlike the case c), this
time only two eigenvalues meet and again divide after a while and no longer interact
with the other eigenvalues. It is also apparent from other numerical calculations that
only one pair of complex eigenvalues should appear simultaneously. The eigenvectors
take a complex form similar to the one in case b)
Ψ =
A cos(k+x) + k−k+ sin(ak−) sin(ak+)−α2 cos(ak+) sin(ak−) tan(ak−)k−k+ cos(ak+) sin(ak−)−α2 cos(ak−) sin(ak+)
A sin(k+x)iα sin(ak+)(k−k+ sin(ak+)−α2 cos(ak+) tan(ak−))k−(k−k+ cos(ak+) sin(ak−)−α2 cos(ak−) sin(ak+))
A cos(k−x)− iα cos(ak+)k− cos(ak−)A sin(k−x)
.
39
0 1 2 3 4 5 6Α
5
10
15
20
25
Re Λ
1 2 3 4 5 6Α
-6
-4
-2
2
4
6
Im Λ
Figure 6.6: α-dependence of eigenvalues for c = 0.4, a = π4 , β = −0.5 in Example d).
6.2 Metric operator
In the end of this chapter we will take a look on the metric operator Θα for our spinor
model as it was introduced in Section 2.2. We will consider only a simple example of
the boundary conditions (5.2) with the matrix
A =
(iα 0
0 iα
).
We will make a use of the knowledge of metric operator θα for one dimensional case with
the Hamiltonian h acting in the Hilbert space L2((−a, a))
hψ := −ψ′′,D(h) :=
{ψ ∈ H2((−a, a))
∣∣ψ′(±a) + iαψ(±a) = 0},
where ψ ∈ L2((−a, a)) and α ∈ R. The formula for the metric related to this Hamiltonian
was firstly found in [21] and the elegant way of expressing this metric as
θα = I +K, (6.10)
where K is an integral operator with kernel
K (x, y) =1
2a
(eiα(y−x) − 1
)+
iα
2a(y − x)− α2
2axy +
α2a
2+
(iα− α2
2(x− y)
)sgn(x− y),
(6.11)
40
was found in [32]. With the knowledge of eigenvalues (6.8) we can simplify the corre-
sponding eigenvectors to the form
Ψ+n =
(cos(k+
n x)− i αk+n
sin(k+n x)
0
)
Ψ−n =
(0
cos(k−n x)− i αk−n
sin(k−n x)
),
(6.12)
where k±n =√λj,± ± c. These vectors corresponding to different eigenvalues are clearly
independent and therefore we can with the methods used in [21] find the metric expressed
using the functions
φ±0 =
√1
2aeiαx,
φ±n =
√1
2a
(cos(k±n x) + i
α
k±nsin(k±n x)
), n ≥ 1.
The metric then takes the form of strongly convergent series
Θα =∞∑n=0
(φ+n
0
)((φ+n
0
), ·)
+∞∑n=0
(0
φ−n
)((0
φ−n
), ·)
=
(θ 0
0 0
)+
(0 0
0 θα
)
= θα
(1 0
0 1
),
(6.13)
where (·, ·) is the scalar product (5.1) and θα is the one-dimensional metric (6.10). The
metric Θα is bounded, symmetric, non-negative and satisfies the relation H∗ΘαΨ =
ΘαHΨ for all Ψ ∈ D(H). If we assume α /∈ nπ2a ± c) with n ∈ Z \ {0} then the metric is
positive. This can be immediately proven using the fact that these properties are already
known for the one-dimensional metric θα in [21] and [32] (positivity). Peculiar feature
of this metric is its form as a multiplication of the identity matrix and its dependence
on the constant c. This constant does not appear in the explicit formula of the metric
but only in the requirement on the positivity of the metric.
41
Chapter 7
Conclusions
In this thesis we investigated the influence of the magnetic field with Robin-type bound-
ary conditions on a charged particle. The physically motivated Hamiltonian was intro-
duced in a rigorous mathematical way via the quadratic form and further investigated.
We proved that it is an m-sectorial operator with a compact resolvent. Main tool used
during the investigation was the First representation theorem which was applied in this
case of two dimensional vectors. The quadratic form approach enabled us to find easily
the adjoint operator and conditions on self-adjointness, P-symmetry, T -symmetry and
PT -symmetry of the Hamiltonian, which restricted the boundary conditions.
Many types of boundary conditions were studied and few examples with purely real
spectrum were found. Among these, there is an interesting self-adjoint example that
connects both parts of the spinor. The numerical simulations suggest that if complex
eigenvalues appear in the spectrum, there will always be simultaneously only one pair
of complex conjugated eigenvalues. And finally for one of these boundary conditions we
were able to find a metric operator which is just a multiplication of the identity matrix
with the one-dimensional metric.
There still remain few open questions for further research. The list of the boundary
conditions were not complete since we considered in this thesis only the PT -symmetric
boundary conditions. There are still undiscovered possibilities for PT -symmetric bound-
ary conditions with real spectrum and finding corresponding metric operator. At least
but not last this thesis opened the way to further multidimensional generalisations of
the original simple PT -symmetric model.
42
Appendix A
Wolfram Mathematica source
code
We state the source code of the Wolfram Mathematice software which was used for
numerical calculations of α-dependencies of the points of the spectrum in Chapter 6 in
Example d).
f [ \ [ Lambda ] , \ [ Alpha ] , a , c ] := 4 \ [ Alpha ]ˆ2 Sqrt[−c +\[Lambda
] ]
Sqrt [ c +\[Lambda ] ] Cos [ a Sqrt [ c +\[Lambda ] ] ] ˆ 2 Sin [ a Sqrt[−c +\[Lambda ] ] ] ˆ 2
+4 \ [ Alpha ]ˆ2 Sqrt[−c +\[Lambda ] ] Sqrt [ c +\[Lambda ] ]
Cos [ a Sqrt[−c +\[Lambda ] ] ] ˆ 2 Sin [ a Sqrt [ c +\[Lambda ] ] ] ˆ 2
+(c ˆ2−\[Alpha ]ˆ4−\ [Lambda ] ˆ 2 ) Sin [ 2 a Sqrt[−c +\[Lambda ] ] ]
Sin [ 2 a Sqrt [ c +\[Lambda ] ] ]
Clear [ eps , imax ,m, a , alpha ,w, w01 , w1 , w02 , w2 , w03 , w3 , w04 , w4 , r , s ] ;
eps =0.001; imax=500;a =3.14/4; c =0.4 ; r =6; s =25;
w1=0.7; w2=3.6; w3=4.4; w4=15.5 ;
Table [ a lpha [ i ] ,{ i , 0 , imax } ] ; Table [ w01 [ i ] ,{ i , 0 , imax } ] ; Table [ w02 [ i
] ,{ i , 0 , imax } ] ;
Table [ w03 [ i ] ,{ i , 0 , imax } ] ; Table [ w04 [ i ] ,{ i , 0 , imax } ] ;
For [ i =0, i<imax+1, i ++, alpha [ i ]=0.5+ r ∗ i / imax ;
w1=w/ .FindRoot [ f [w, alpha [ i ] , a , c ]==0,{w, w1 } ] ;
w2=w/ .FindRoot [ f [w, alpha [ i ] , a , c ]==0,{w, w2 } ] ;
w3=w/ .FindRoot [ f [w, alpha [ i ] , a , c ]==0,{w, w3 } ] ;
w4=w/ .FindRoot [ f [w, alpha [ i ] , a , c ]==0,{w, w4 } ] ;
43
w01 [ i ]=w1 ; w02 [ i ]=w2 ; w03 [ i ]=w3 ; w04 [ i ]=w4 ;
w1=I f [Abs [Re [ w01 [ i ]−w02 [ i ] ] ] < eps && Abs [Im [ w01 [ i ]−w02 [ i ] ] ] < eps ,
w1+0.5∗I , w1 ] ;
w2=I f [Abs [Re [ w01 [ i ]−w02 [ i ] ] ] < eps && Abs [Im [ w01 [ i ]−w02 [ i ] ] ] < eps ,
w2−0.5∗I , w2 ] ;
w1=I f [Abs [Re [ w01 [ i ]−w02 [ i ] ] ] < eps && Abs [Im [ w01 [ i ]−w02 [ i ] ] ] < eps ,
w1+0.5 ,w1 ] ;
w3=I f [Abs [Re [ w03 [ i ]−w04 [ i ] ] ] < eps && Abs [Im [ w03 [ i ]−w04 [ i ] ] ] < eps ,
w3+0.5∗I , w3 ] ;
w4=I f [Abs [Re [ w03 [ i ]−w04 [ i ] ] ] < eps && Abs [Im [ w03 [ i ]−w04 [ i ] ] ] < eps ,
w4−0.5∗I , w4 ] ;
w3=I f [Abs [Re [ w03 [ i ]−w04 [ i ] ] ] < eps && Abs [Im [ w03 [ i ]−w04 [ i ] ] ] <500
eps , w3+0.5 ,w3 ] ]
Plot [{Interpolation [ Table [{ alpha [ i ] ,Re [ w01 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Re [ w02 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Re [ w03 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Re [ w04 [ i ] ] } , { i , 0 , imax } ] ] [ x ]} ,
{x , 0 . 5 , r } , PlotStyle−>{{Blue} ,{Blue} ,{Blue} ,{Blue}} ,
AxesLabel−>{Sty l e [ \ [ Alpha ] , 1 6 ] , S ty l e [ ”Re\ [ Lambda ] ” , 1 6 ]} ,
PlotRange−>{0, s } ,AxesOrigin−>{0 ,0}]Plot [{Interpolation [ Table [{ alpha [ i ] ,Im [ w01 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Im [ w02 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Im [ w03 [ i ] ] } , { i , 0 , imax } ] ] [ x ] ,
Interpolation [ Table [{ alpha [ i ] ,Im [ w04 [ i ] ] } , { i , 0 , imax } ] ] [ x ]} ,
{x , 0 . 5 , r } , PlotStyle−>{{Red, Dashing [ Medium ]} ,{Red, Dashing [ Medium
]} ,
{Red, Dashing [ Medium ]} ,{Red, Dashing [ Medium ]}} , AxesLabel−>{Sty l e
[ \ [ Alpha ] , 1 6 ] ,
S ty l e [ ”Im\ [ Lambda ] ” , 1 6 ]} , PlotRange−>{−6,6},AxesOrigin−>{0 ,0}]
44
Bibliography
[1] Adams, R. A. Sobolev spaces. Academic Press, New York, 1975.
[2] Albeverio, S., Fei, S. M., and Kurasov, P. Point Interactions: PT-Hermiticity
and Reality of the Spectrum. Letters in Mathematical Physics 59 (2002), 227–242.
[3] Albeverio, S., Gunther, U., and Kuzhel, S. J-self-adjoint operators with
C-symmetries: an extension theory approach. Journal of Physics A: Mathematical
and Theoretical 42 (2009), 105205 (22pp).
[4] Albeverio, S., and Kuzhel, S. One-dimensional Schrodinger operators with P-
symmetric zero-range potentials. Journal of Physics A: Mathematical and General
38 (2005), 4975–4988.
[5] Bender, C. M., and Boettcher, S. Real Spectra in Non-Hermitian Hamiltoni-
ans Having PT Symmetry. Physical Review Letters 80 (1998), 5243–5246.
[6] Bender, C. M., and Tan, B. Calculation of the hidden symmetry operator for
a PT -symmetric square well. Journal of Physics A: Mathematical and General 39
(2006), 1945.
[7] Borisov, D., and Krejcirık, D. PT-symmetric waveguides. Integral Equations
Operator Theory 62 (2008), 489–515.
[8] Borisov, D., and Krejcirık, D. The effective Hamiltonian for thin layers with
non-Hermitian Robin-type boundary conditions. Asymptotic Analysis (to appear).
[9] Dieudonne, J. Quasi-Hermitian operators. Proceedings Of The International
Symposium on Linear Spaces (July 1961), 115–123.
[10] Edmunds, D. E., and Evans, W. D. Spectral Theory and Differential Operators
(Oxford Mathematical Monographs). Oxford University Press, USA, 1987.
45
[11] Emil Prodan, S. R. G., and Putinar, M. Norm estimates of complex symmetric
operators applied to quantum systems. Journal of Physics A: Mathematical and
General 39, 2 (2006).
[12] Formanek, J. Uvod do kvantove teorie I.,II. Academia, 2004.
[13] Garcia, S. R., and Putinar, M. Complex symmetric operators and applications.
Transactions Of The American Mathematical Society 358 (2006), 1285–1315.
[14] Garcia, S. R., and Putinar, M. Complex symmetric operators and applications
II. Transactions Of The American Mathematical Society 359 (2007), 3913–3931.
[15] Griffiths, D. Introduction to Electrodynamics. 1999.
[16] Griffiths, D. Introduction to Quantum Mechanics, 2nd edition ed. 2005.
[17] Hernandez-Coronado, H., Krejcirık, D., and Siegl, P. Perfect transmission
scattering as a PT -symmetric spectral problem. Physics Letters A 375 (2011),
21492152.
[18] Hlavaty, L. Slabikar kvantove mechaniky. electronic textbook, September 2009.
[19] Kato, T. Perturbation theory for linear operators. Springer-Verlag, 1966.
[20] Krejcirık, D. Calculation of the metric in the Hilbert space of a PT -symmetric
model via the spectral theorem. Journal of Physics A: Mathematical and Theoretical
41 (2008), 244012.
[21] Krejcirık, D., Bıla, H., and Znojil, M. Closed formula for the metric in the
Hilbert space of a PT -symmetric model. Journal of Physics A: Mathematical and
General 39 (2006), 10143–10153.
[22] Krejcirık, D., and Siegl, P. PT -symmetric models in curved manifolds. Jour-
nal of Physics A: Mathematical and Theoretical 43, 48 (2010), 485204.
[23] Krejcirık, D., and Tater, M. Non-Hermitian spectral effects in a PT -symmetric
waveguide. Journal of Physics A: Mathematical and Theoretical 41 (2008), 244013
(14pp).
[24] Langer, H., and Tretter, C. A Krein Space Approach to PT-symmetry.
Czechoslovak Journal of Physics 54 (2004), 1113–1120.
46
[25] Mostafazadeh, A. Pseudo-Hermiticity versus PT-symmetry I: The necessary
condition for the reality of the spectrum of a non-Hermitian Hamiltonian. Journal
of Mathematical Physics 43 (2002), 205–214.
[26] Mostafazadeh, A. Metric operators for quasi-Hermitian Hamiltonians and sym-
metries of equivalent Hermitian Hamiltonians. Journal of Physics A: Mathematical
and Theoretical 41 (2008), 244017.
[27] Mostafazadeh, A., and Batal, A. Physical aspects of pseudo-Hermitian and PT
-symmetric quantum mechanics. Journal of Physics A: Mathematical and General
37 (2004), 11645.
[28] Musumbu, G., and Heiss. Choice of a metric for the non-Hermitian oscillator.
Journal of Physics A: Mathematical and Theoretical 40, 2 (2007).
[29] Scholtz, F., and Geyer, H. Operator equations and Moyal products-metrics in
quasi-Hermitian quantum mechanics. Physics Letters B 634, 1 (2006), 84 – 92.
[30] Scholtz, F. G., Geyer, H. B., and Hahne, F. J. W. Quasi-Hermitian oper-
ators in quantum mechanics and the variational principle. Annals of Physics 213
(1992), 74–101.
[31] Siegl, P. Quasi-Hermitian Models. Master’s thesis, FNSPE, CTU in Prague,
2007/2008.
[32] Zelezny, J. The Krein-space theory for non-Hermitian PT -symmetric operators.
Master’s thesis, FNSPE, CTU in Prague, 2010/2011.
[33] Zelezny, J. Spectrum of the metric operator of a simple PT -symmetric model.
International Journal of Theoretical Physics 50 (2011), 1012–1018.
47